Because humans still get the picture selection prompts, so you'd have to write code good enough to read the prompt and then select the appropriate images.